Integrated Power Management for CPU and GPU Based on Prediction of the Influence of Mutual Dependency between Processing Units

Modern mobile devices are commonly equipped with a system-on-chip (SoC) which commonly integrates a central processing unit (CPU), a graphic processing unit (GPU), and digital signal processors (DSPs). Among them, CPU and GPU consume large amount of power. There have been various researches about power management of CPUs and GPUs. However, most of them studied power management only for the CPU or the GPU by treating them as independent devices. There exists mutual dependency between the CPU and the GPU. When the clock speed of either the CPU or the GPU is changed, the otherâ?™s desirable clock speed may also be changed. Therefore, it is necessary to consider the mutual dependency between the CPU and the GPU to effectively manage the states of the CPU and the GPU for low power. In this paper, we propose a method for integrated power management for the CPU and the GPU by predicting the influence of mutual dependency between the processing units. When compared to a conventional independent management scheme, the power consumption is reduced by about 8% while the benchmark score remains almost unchanged. Thus, the proposed scheme increases power efficiency by about 10%.